A hallucination is when an AI states something false as if it were fact, such as a candidate's past employer, a law that does not exist or a made up source. The text reads confidently, which is why errors slip through unless a person checks names, numbers and claims.
AItopic
2 minread
2related terms
In one line
When an AI states something false as fact, such as a candidate's past employer, a law that does not exist or a made up source. Always check names, numbers and claims before they reach a candidate or client.
Generative AI predicts likely words rather than looking facts up, so when it lacks information it may fill the gap with something that sounds right. Hallucinations are more common when you ask about niche topics, recent events, specific people or exact figures.
In recruiting the risk shows up in candidate summaries that add skills nobody mentioned, outreach that praises a project the person never worked on, salary figures with no source and compliance answers that cite the wrong rule. Each one can embarrass you with a candidate or client, or worse, lead to a bad decision.
Tools reduce hallucinations by grounding answers in trusted data, such as your ATS records or an uploaded document, and showing where each claim came from. Grounding helps, but it does not remove the need for a human check.
Why it matters to recruiters
Recruiters trade on credibility. One invented detail in a candidate summary sent to a client, or a wrong legal claim in an HR note, can cost a placement or create real risk. Treating AI output as a draft that needs checking protects your reputation and keeps decisions about people based on facts.
An example from a recruiting desk
After a candidate call, an agency recruiter asks an AI assistant for a summary and a submittal. The draft says the candidate led a team of 12 at a previous employer. The call notes say she mentored two junior engineers. The recruiter fixes the line before sending. Had the client asked about it in interview, the error would have cost trust in both of them.
Tips that help
Ask the AI to quote or point to the source for every factual claim.
Check names, dates, titles, numbers and legal claims against the original record.
Use tools that ground answers in your own documents for candidate summaries.
Tell the AI it may answer 'I do not know' when information is missing.
Mistakes to avoid
Trusting a confident tone as a sign the content is accurate.
Letting AI written summaries go to clients without comparing them with your notes.
Frequently asked questions
Why does AI hallucinate?
Language models generate likely sounding text from patterns, not from a verified database. When the answer is not in what they were given or learned, they can produce something plausible but wrong instead of saying they do not know.
Can hallucinations be removed completely?
Not today. Grounding, better prompts and source citations reduce them a lot, but no vendor can promise zero errors. Keep a human review step on anything that reaches a candidate, client or hiring decision.
How do I spot a hallucination quickly?
Look for specific details you did not supply, such as numbers, titles, dates or citations. Those are the most likely to be invented. Check each against the source before you use it.
